Automatic segmentation of plant leaves disease using min-max hue histogram and k-mean clustering

Automatic segmentation of plant image's leaf diseases has recently become a popular area of study worldwide. The suggested approach automatically segments various areas of leaf disease from images of the plant, which can then be combined with machine learning or deep learning techniques to improve system accuracy. Our suggested method consists of three stages: Preprocessing is applied in the first stage where a rank order fuzzy (ROF) filter is proposed that reduces the background noise from the plant picture. In the next stage, disease spot detection is performed using proposed min-max hue histogram based techniques. Disease spot identification prior to segmentation helps in proper segmentation of k-mean clustering. The K-means clustering is then performed in the next stage to segment the leaf pictures into uniform regions. These segments are transformed into HSI color spaces and the segment with the largest hue value is extracted as the disease segment. The proposed methodology is implemented in Matlab 18a and studies are carried out on various plant images. The proposed ROF filter demonstrates superior results to the other state-of-the-art filters. The filter is also resistant to very large noise levels, and shows meaningful details at noise levels of 95%. Besides, our hue-based spot detection is compared with the existing method and it can be shown by the suggested approach, the diseases have been found mostly correctly. The segmentation accuracy of the proposed method is calculated using the Jaccard coefficient, Sensitivity and Positive Prediction Rate. Our proposed system achieved high Jaccard coefficient value of 0.7747.

Automated summary

Automatic segmentation of plant image's leaf diseases has become a popular area of study worldwide. The proposed method utilizes machine learning and image processing techniques to automatically segment disease regions and improve accuracy by comparing hue values in different color spaces.